Papers CHI **** ** MARCH * APRIL
Doom as an Interface for Process Management
Dennis Chao
Computer Science Department
University of New Mexico
Albuquerque, NM 87131 USA
******@**.***.***
ABSTRACT
This paper explores a novel interface to a system adminis-
tration task. Instead of creating an interface de novo for the
task, the author modi ed a popular computer game, Doom,
to perform useful work. The game was chosen for its appeal
to the target audience of system administrators. The imple-
mentation described is not a mature application, but it illus-
trates important points about user interfaces and our relation-
ship with computers. The application relies on a computer
game vernacular rather than the simulations of physical re-
ality found in typical navigable virtual environments. Using
a computer game vocabulary may broaden an application s
audience by providing an intuitive environment for children
Figure 1: The PSDoom interface.
and non-technical users. In addition, the application high-
lights the adversarial relationships that exist in a computer
and suggests a new resource allocation scheme.
a rst-person shooter game. Each process can be represented
as a monster, and interacting with the monsters would af-
Keywords
fect the corresponding processes. The implementation of this
Cyberspace, Doom, rst-person shooter, games, metaphors,
metaphor and the great interest it generated reveal interesting
operating systems, Post-Modernism, 3D user interfaces, ver-
insights about our computers and our society.
nacular, video games, visualization
IMPLEMENTATION
INTRODUCTION
Those who use computers inevitably encounter some of the The Doom process manager (PSDoom) is a modi cation of
metaphors that allow for easier assimilation of abstract con- the game Doom [8] that displays representations of the pro-
cepts. The desktop metaphor is so pervasive that most users cesses running on a machine. Rather than using standard
hardly notice it [9], but the richness of the science- ction ver- text-mode UNIX tools to view and manipulate processes, one
sion of cyberspace is largely con ned to research laboratories surveys and shoots at a room full of bloodthirsty mutants, as
and Hollywood. The application described in this paper is an shown in Figure 1. When a user starts PSDoom, currently
initial step towards bringing a richer environment to personal running processes are instantiated as process monsters in a
computers. single room in a dungeon. These monsters have their asso-
ciated process name and id printed on them. The program
There is a large gap between how we think about perform- periodically polls the operating system to add newly-created
ing actions on our computers and how we actually perform processes to the game. The user may choose to view the pro-
them. For example, people who need to manage processes cesses from a balcony above the room, as shown in Figure 2,
on a UNIX system think about the daemons spawning chil- or to enter the room to interact with them. If the user in icts
dren that may need to be killed or blown away. This vio- a wound upon a process monster, the corresponding process
lent language suggests a metaphor for process management: priority is lowered to give it fewer CPU cycles. When the
monster accumulates enough damage and is killed, the asso-
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies ciated process is also killed.
are not made or distributed for pro t or commercial advantage and that
copies bear this notice and the full citation on the rst page. To copy
PSDoom inherits the rest of its behavior from the original
otherwise, or republish, to post on servers or to redistribute to lists,
Doom, and play is not noticeably affected. Monsters attempt
requires prior speci c permission and/or a fee.
to attack the player and each other. The hostility of the mon-
SIGCHI 01, March 31-April 4, 2001, Seattle, WA, USA.
Copyright 2001 ACM 1-58113-327-8/01/0003. . . $5.00.
CHI 2001
152 Volume No. 3, Issue No. 1
CHI 2001 31 MARCH 5 APRIL Papers
Figure 2: The view from the balcony in PSDoom. Figure 3: Killing a process in PSDoom.
sters and the user s limited ammunition are disincentives to Doom is that monsters are much more likely to attack each
attack them. Con ict among process monsters could help other than expected. This causes many windows to mysteri-
regulate heavily-utilized systems by making crowded rooms ously disappear as the program runs. For the same reason,
have higher mortality rates. Killing random processes on an the computer is prone to crashing because certain processes
extremely loaded system is not an uncommon operating sys- are vital to the computer s operation and should not be killed.
tem strategy. When the user is killed, he or she will be
Many users want a larger variety of monsters in PSDoom. If
healed and placed at the entrance of the dungeon with a pis-
larger or more important processes were represented as larger
tol and a modest amount of ammunition.
monsters, it would be easier to assess the machine load at
a glance. If these monsters were also more powerful, they
Doom was chosen for this project for two reasons. The rst is
would be less likely to be killed by accident and be more able
that it is a classic game, familiar to most system administra-
to defend themselves against the player. Several users made
tors. The second is that its source code was recently released
similar suggestions for altering the appearance of monsters
under the GNU General Public License (GPL) [13]. This li-
based on certain attributes. Processes that take more memory
cense not only allows the author to modify the source code,
could appear wider, while those that take more CPU time
but it guarantees that future derivatives of the author s work
can appear taller. Sleeping processes could be represented
will be available to the public.
by napping monsters.
RESULTS
To address some of these requests, I added code to make
PSDoom received a surprisingly large reaction even though
some of the more important processes Barons of Hell, the
it was not publicized [24, 26, 4, 17]. Less than a week af-
largest monsters in the game. Unfortunately, they had a ten-
ter the initial version of the code was written, the project s
dency to quickly kill all of the other processes, and the user
website was attracting tens of thousands of visitors per day.
could not interact with processes for more than a few seconds
Approximately 800 responses were e-mailed to the author or
before his or her avatar is killed. Making the monsters less
www.slashdot.org within the rst two months. Of these re-
aggressive would allow the user to navigate among processes
sponses 27% praised the project, 23% offered suggestions
more easily as well as make the computer more stable.
for improving PSDoom, 10% found the project funny, 10%
reported technical problems, 8% related PSDoom to science
DISCUSSION
ction or to the future of interfaces, 1% disliked the project,
The enormous interest that PSDoom generated naturally raises
and 0.6% were frightened by its implications.
the question of why people nd it so compelling. Perhaps
even more interesting than the application itself is the set of
Users found the interface intuitive. One can quickly assess
issues that it raises.
machine load by seeing how crowded a room is. The com-
mand line methods to slow down and kill processes are dif-
Cyberspace Environments
ferent, while PSDoom uni es them shooting a monster
In 1981, Vernor Vinge introduced the concept of cyberspace
with a small weapon slows down or wounds the corre-
to the reading public in his novella True Names [30], in which
sponding process, and repeated rings or the use of a large
characters could plug into a virtual universe where their ac-
weapon kills the process, as shown in Figure 3. The violence
tions in a fantasy world mapped to performing sophisticated
in icted upon the monsters re ects the violent terminology
actions on the network. For example, navigating a treacher-
of UNIX commands.
ous path through a swamp in cyberspace could gain the user
A signi cant problem with the current implementation of PS- entry into a high-security network in the real world.
anyone. anywhere. 153
Papers CHI 2001 31 MARCH 5 APRIL
The use of a cyberspace environment can make performing
complex tasks easy if the mapping between the actions in the
virtual world and their effects in the real world is intuitive.
Communication among users may be facilitated because in-
teracting with a group of avatars can be made as natural as
interacting with a group of people. Doom places the user
into a 3D rst-person perspective environment that rapidly
engages the senses. Its interface is particularly suitable for
the task of process management because many system ad-
ministrators are familiar with the game and it realizes the
common metaphor of killing processes.
In PSDoom, actions that affect the processes take time and
effort. This philosophy runs counter to intuition, which tells
Figure 4: General Magic s MagicCap.
us to make everything as easy to do as possible. In a com-
mand line interface, all actions take approximately the same
amount of work, making it just as easy to erase all of one s
les as it is to remove a single one. In this cyberspace envi-
ronment, the players are not omnipotent, so actions take an
amount of effort proportional to their effects. People with
different levels of authority can be given weapons of differ-
ent strengths. An experienced system administrator can be
given a large gun, while the beginner may be forced to deal
with monsters with his or her bare hands. It would take a
foolhardy player to attack a room full of monsters, just as
a newbie should not kill a bunch of important processes. A
more experienced sysadmin would have time to stop a novice
who is trying to kill the wrong process.
The Software Vernacular
Current applications often do not leverage the rich vocabu-
lary of contemporary mass media and video games. Their
Figure 5: Microsoft Bob.
user interfaces are usually austere, re ecting the limitations
of the machines of two decades ago. Our rectilinear desktops
feature a Machine Aesthetic, not a human one. There have sentation and artistic expression [6]. The idioms of cartoons
been attempts to create friendlier interfaces based on familiar form an effective vocabulary for concise visual communica-
physical settings (e.g. General Magic s MagicCap [21] (Fig- tion with the user, but cartoons lack a crucial element of user
ure 4) and Microsoft Bob [2] (Figure 5)), but they have often interfaces: interactivity. Computer games have successfully
been derided as condescending or con ning [12]. The failure integrated concise, stylized visual representations and feed-
of these more humane interfaces may be due to the denial back from the user.
of the computer world s own vocabulary that is distinct from
that of everyday life s. In many ways, computer games have The desktop metaphor derives much of its power from the
become the vernacular of the computer world because they fact that users were traditionally of ce workers who spent
are what children and many non-technical adults use. much of their days at actual desks [12, 20]. However, in re-
cent years computers have entered the home, and the current
Interfaces that adhere strictly to a metaphor based on the generation of users is gaining increased exposure to comput-
physical world are in danger of becoming simulations of ers that are not merely of ce appliances. Children are now
reality instead of useful analogies [18]. One problem with growing up on MTV and Nintendo and are therefore more lit-
simulations is that one must closely adhere to all the restric- erate in the languages of video games and mass media than
tions of the real system. For example, in Microsoft Bob it is in that of the traditional of ce milieu. Designers should con-
disturbing to see pieces of furniture hovering over the oor sider taking advantage of the vocabulary of popular culture
in apparent violation of gravity. In contrast, it is more or instead of trapping the next generation in dioramas of subur-
less natural to have an arbitrarily large number of windows ban homes and of ces.
open on a computer desktop, which is a less-literal, and thus
less-con ning, metaphor. Cartoon-like animations have been Like the Post-Modern architects, I would like to blur the
suggested as a compromise between straightforward repre- distinction between high and low culture [29, 3] and
CHI 2001
154 Volume No. 3, Issue No. 1
CHI 2001 31 MARCH 5 APRIL Papers
bring popular culture into serious computer applications.
By adding familiar cultural elements to our interfaces, we
can make them appealing to the common user and children,
the next generation of users. These elements would bring a
sense of playfulness and ease that are usually absent from our
button- and toolbar- laden windows.
Several attempts have been made to introduce game-like el-
ements to the desktop, including the recent UBUBU Uni-
verse [27]. UBUBU replaces the hyperlinks and folders on
the desktop with navigable planets in a solar system. While
the concept is super cially similar to Bob, the choice of y-
ing through space rather than walking around a house may
engage the general public instead of being perceived as pa-
Figure 6: The UNIX top application.
tronizing. Companies like UBUBU that design products for
web-surfers are using playful game-like designs to tap into
the lucrative demographic of computer game players. avatar loses possession of the items he or she was carrying
and is sent home. About half the players enjoyed this aspect
An Adversarial Operating System
of the environment while the rest were opposed to it [23]. To
PSDoom presents an unusual perspective on our interactions
stem anarchy, the players established their own churches and
with computer programs. It is unique in that it allows the
laws. In Ultima Online [28], an online fantasy environment,
processes to ght each other and the user. Thinking of our
player-killing is an integral part of gameplay for some, but
computing environments as being adversarial can be enlight-
many value the complex social interactions with their fellow
ening. Processes are competing for machine resources, such
players even more.
as memory and CPU time, and PSDoom makes this compe-
tition explicit. The user may want to kill processes to free We have seen violent multi-user games evolve into peace-
needed resources, so from the process s perspective, the user ful virtual communities. Production values play a role in the
may be its greatest threat. The processes are given the ability suspension of disbelief essential for this to happen [19], but
to shoot back and defend themselves. even text-based MUDs, which originated as hack-and-slash
games, have become chatrooms [5]. The key elements seem
Operating systems attempt to satisfy users and processes
to be the rich interactions with other players and the persis-
through fairness policies for resource allocation. An adver-
tence of identity over multiple sessions. These seminal ex-
sarial operating system would invert the responsibility for
periments in multi-user virtual worlds are showing us how to
sharing resources by forcing the users and processes to com-
moderate violence with a sense of community.
pete directly with each other. It would simply maintain a
stable arena in which processes could compete. An offensive
RELATED WORK
and defensive arms race could ensue, but the system would
Visualizing Processes
not necessarily degrade to an hostile, anarchic environment.
The standard UNIX application to view processes is top (Fig-
Under the right conditions, processes will cooperate rather
ure 6). It is a dynamic version of ps, a command that simply
than compete [1]. They would nd it in their best interests to
lists the running processes. As implied by its name, top dis-
only attack when necessary because unduly belligerent pro-
plays the top CPU-users, refreshing at regular intervals. Its
cesses would become injured in frequent con icts and die.
interface is primitive. The user may not interact directly with
the process list an instruction and process number must be
Violence in Computer Games
entered, much like on the command line. Another limitation
It is unfortunate, especially in the light of recent schoolyard
is that the number of processes displayed is limited by the
tragedies, that rst-person shooters are so popular. Even
number of lines on the display (typically around twenty to
though studies on the effects of violent media on youth are
thirty). There is now a plethora of alternatives to top that
not conclusive [11], interface designers must proceed with
allow more interaction with the list, such as clicking on the
caution when adding potentially aggressive aspects to inter-
entries. These contribute obvious graphical additions to a
faces. One must take into consideration the age of the ex-
process display and add little conceptually.
pected user base and the possible use of non-aggressive al-
ternatives.
LavaPS [16] is a more abstract way to view processes. It
Lucas lm s Habitat [14], an early multi-user virtual envi- displays processes as colored blobs in a window, using the
ronment, faced these ethical dilemmas with the introduction size, color, and location of shapes in the window to repre-
of weapons. These weapons could kill avatars, though sent certain metrics of the processes, as shown in Figure 7.
death in Habitat is no more serious than death in Doom the More explicit information can be shown on demand. Like
anyone. anywhere. 155
Papers CHI 2001 31 MARCH 5 APRIL
affects ease-of-use, may be desirable because it can make the
task more enjoyable [22].
The use of game-like interfaces need not harm the perfor-
mance of applications. In fact, it can reduce the learning
curve. We would not need to sacri ce power games have
amazingly complex maneuvers that skilled operators can use
in addition to their simple repertoires for novices.
FUTURE WORK
It would not be dif cult to remedy the major source of in-
stability of PSDoom, the belligerence of the monsters. The
monsters currently have no sense of self-preservation. They
will continue to attack until they or their targets are elimi-
nated. One could simply make the monsters only attack when
Figure 7: Heidemann s LavaPS process visualization ap-
plication. provoked and more likely to run away rather than ght to the
end.
PSDoom, LavaPS represents processes as physical entities It would be natural to extend the program to networked envi-
with no quantitative information. However, the application ronments. In fact, many internet search engines have already
is inspired by calm technology [31] and is therefore in- classi ed PSDoom as a network tool in anticipation of this
tentionally passive. Its main purpose is to display the state feature. A networked version would allow multiple users to
of processes, and manipulating the actual processes can be communicate verbally as well as work cooperatively [10].
done through a menu. Turning this single-player game environment into an online
world could reduce the level of violence by increasing the
Game-like Interfaces sense of community.
The CHI 97 Workshop on Game Design and HCI suggests
By selecting only certain types of processes to appear in the
that computer game design has much to offer to the HCI com-
dungeon, one can change the nature of PSDoom. For in-
munity, but little has been done so far [7]. A popular area for
stance a lter that only allows suspicious or possibly cor-
game-like interfaces is software for children. Such products
rupted processes [25] to appear would turn PSDoom into a
range from educational games to tools that encourage cre-
security program. Alternatively, one can choose to display
ativity, such as authoring tools. It is dif cult to make direct
only large users of memory or bandwidth in order to allevi-
comparisons between software for children and software for
ate pressure on these resources.
adults because the expectations we have for the intended au-
diences are so different. Designers readily use entertaining
CONCLUSION
interfaces to keep children engaged [15]. Even in educational
PSDoom presents a novel approach to process visualization
software, in which the ultimate goal is to teach the user, the
and manipulation that illuminates many aspects of our rela-
proximate goal is to keep the child engaged, or else the ul-
tionship with computers. The popularity of this early pro-
timate goal will never be reached. In application software
totype demonstrates that interfaces inspired by video games
for adults, the burden of staying focused is on the user. The
and science ction are eagerly awaited by a large audience.
measurement for success is not how enjoyable the software
Computer games are a rich and often overlooked source of
is but rather how effectively it accomplishes a task.
technology and metaphors for user interfaces that will only
increase in importance as children, the next generation of
The contrast between software for these two audiences high-
users, grow up with computers. Our computer environments
lights the difference between enjoyment and ergonomics. A
are rapidly becoming an anachronism as people who have
piece of children s software must look good and be fun, while
never handled actual le folders are forced to use their virtual
software for adults must work. These are not necessarily con-
counterparts. The process of changing interface metaphors
icting goals, as an interface that is dif cult to use will also
will expose previously-obscured facets of the underlying ap-
adversely affect the user s enjoyment. However, for seri-
plications, such as the genuinely adversarial interactions in
ous software, concessions to make a program more enjoy-
our operating systems. Interested programmers can explore
able might not be implemented because they can not be jus-
and modify the code for PSDoom, which is available at
ti ed on ergonomic grounds.
http://www.cs.unm.edu/ dlchao/ ake/doom.
For an application that helps the user monitor the state of a
system, like PSDoom, the case for making user interfaces ACKNOWLEDGMENTS
more engaging is compelling. Keeping the user engaged is I would like to thank the Adaptive Computation Group at
as important as making the tasks easy to perform. In such UNM for providing a supportive environment in which one
cases, the use of randomness in a program, which adversely can claim one is doing research while playing Doom for
CHI 2001
154 Volume No. 3, Issue No. 1
156
CHI 2001 31 MARCH 5 APRIL Papers
two days. The offhand remarks of Anil Somayaji provided 14. Habitat. Lucas lm Ltd, San Rafael CA, 1986.
the initial inspiration for the project. David Ackley, Ben
15. Hakansson, J. Lessons Learned From Kids: One Devel-
Bederson, Patrik D haeseleer, Stephanie Forrest, and Jason
oper s Point of View. The Art of Human-Computer Inter-
Stewart provided helpful comments that improved this paper.
face Design, edited by Brenda Laurel. Addison-Wesley
This work was partially supported by the National Science
Publishing Company, Inc., New York, 1990.
Foundation (grants NSF-9553623, CDA-9503064, and IRI-
9711199) and the Of ce of Naval Research (grant N00014- 16. Heidemann, J. LavaPS. 1998. Available at
99-1-0417). http://www.isi.edu/ johnh/SOFTWARE/LAVAPS/.
REFERENCES 17. Hwang, F. Commando Line Interface. Wired 8, 2 (Febru-
1. Axelrod, R. The Evolution of Cooperation. Basic Books, ary 2000), 52.
New York, 1984.
18. Johnson, S. Interface Culture. Harper, San Francisco
2. Bob. Microsoft Corporation, Seattle WA, 1995. CA, 1997.
3. Brand, S. How Buildings Learn: What Happens After 19. Kim, A.J. Ultima Online: An Interactive Virtual World
They re Built. Viking Penguin, New York, 1994. with Multiple Personalities. SIGGRAPH Bulletin 32, 2
(May 1998), 15 19.
4. Brockmeier, J. A SysAdmin s Dream. Linux Magazine
2, 1 (January 2000), 76. 20. Langford, D., and Jones, C. The Kitchen Interface - A
Lateral Approach to GUI. SIGCHI Bulletin 26, 2 (April
5. Bruckman, A. MOOSE Crossing: Construction, Com-
1994), 41 45.
munity, and Learning in a Networked Virtual World for
Kids. MIT Media Lab, Cambridge MA, 1997. 21. MagicCap. General Magic, Sunnyvale CA, 1994.
6. Chang, B.W., and Ungar, D. Animation: From Cartoons 22. Malone, T.W. Toward a theory of intrinsically motivating
to the User Interface. UIST 93 Conference Proceedings instruction. Cognitive Science 4, 333 370.
(Atlanta GA, November 1993), ACM Press, 45 55.
23. Morningstar, C., and Farmer, F.R. The Lessons of Lu-
7. Cherny, L., Clanton, C., and Ostrom, E. Entertainment is cas lm s Habitat. Cyberspace: First Steps, edited by
a Human Factor: A CHI 97 Workshop on Game Design Michael Benedikt. MIT Press, Cambridge MA, 1990,
and HCI. SIGCHI Bulletin 29, 4 (October 1997), 50 54. 273-301.
8. Doom. Id Software, Mesquite TX, 1993. 24. Schmidt, J. Systemverwaltung nach Hackerart. c t:
Magazin fur Computer Technik, 23 (1999), 68.
9. Erickson, T. Working With Interface Metaphors. The Art
of Human-Computer Interface Design, edited by Brenda 25. Somayaji, A. Automated Response Using System-Call
Laurel. Addison-Wesley Publishing Company, Inc., New Delays. 9th USENIX Security Symposium (Denver CO,
York, 1990. August 2000), ACM Press, 185-197.
10. Evard, R. Collaborative Networked Communication: 26. Systemadministration mit dem Colt. IX: Magazin
MUDs as Systems Tools. Proceedings of the Seventh f r professionelle Informationstechnik, 12 (December
u
Systems Administration Conference (LISA VII) (Monter- 1999), 44.
rey CA, November 1993), USENIX, 1 8.
27. UBUBU Universe. UBUBU, Inc., San Francisco CA,
11. The Federal Trade Commission. Marketing Violent En- 2000. Available at http://www.ububu.com/.
tertainment To Children: A Review Of Self-Regulation
And Industry Practices In The Motion Picture, Music 28. Ultima Online. Electronic Arts, Inc., Redwood City CA,
Recording & Electronic Game Industries. Washington: 1997.
Government Printing Of ce, 2000.
29. Venturi, R., Scott Brown, D., and Izenour, S. Learning
12. Gentner, D., and Nielson, J. The Anti-Mac Interface. From Las Vegas. The MIT Press, Cambridge MA, 1972.
Communications of the ACM 39, 8 (August 1996), 70
30. Vinge, V. True Names, in Binary Star #5, edited by
82.
James R. Frenkel. Dell, New York, 1981.
13. GNU General Public License. Free Software Founda-
31. Weiser, M., and Brown, J.S. Designing Calm Technol-
tion, Inc., Boston MA, 1991.
ogy. PowerGrid Journal 1.01 (July 1996).
anyone. anywhere. 157
Copyright 2001 ACM 1-58113-327-8/01/0003. . . $5.00.